We conducted four experiments on the development of motion perception in a total of 109 3- to 5-month-old infants using motion stimuli consisting of opposite-moving dots. A psychophysical study showed that adult subjects perceived two global planes with opposite-moving dots, but this global perception collapsed when paired opposite-moving dots were located within 0.4 deg of one another (Qian, Andersen, & Adelson, 1994). We used this paired-dot stimulus as a non-target and the opponent motion stimulus as a target and examined target preference using methods based on forced-choice-preferential looking (Teller, 1979). In Experiment 1, we used 90 moving dots as stimuli. The results showed that 5-month-old infants had a significant preference for the targets but 4- and 3-month-olds did not. In Experiment 2, we used a small number of dots, and the results showed that 5-month-old infants did not prefer the target significantly. These results suggest that the preference for a target decreases according to the number of dots. In Experiment 3, we used opponent motion with long traveling length of the dots, and the results showed that all age groups, including 3-month-olds, had a preference for the moving targets. We showed that the preference observed in Experiment 3 was dependent not on local traveling length but on the global opponency. These results suggest that the perception of motion opponency based on a global motion cue emerges at 5 months of age (Experiments 1 and 2) and that the traveling length of the dots promote this perception (Experiments 3 and 4).